We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

Transcription

1 We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists 3, , M Open access books available International authors and editors Downloads Our authors are among the 151 Countries delivered to TOP 1% most cited scientists 12.2% Contributors from top 500 universities Selection of our books indexed in the Book Citation Index in Web of Science Core Collection (BKCI) Interested in publishing with us? Contact book.department@intechopen.com Numbers displayed above are based on latest data collected. For more information visit

2 Chapter 2 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going Adrianna Ranger, Navjot Chaudhary and Jonathan Lau Additional information is available at the end of the chapter 1. Introduction In North merica and Europe, cancer involving the central nervous system (CNS) ranks second as the most common malignancy seen in infancy through adolescence, second only to leukaemia [1-7]. Consistent with this are figures on cancer-related mortality. In the year 2004, for example, there were 566 confirmed cases of leukemia-related death among children in the United States, followed by 555 CNS cancer-related deaths; these numbers accounted for 25.5% and 25.0% of the total number of cancer deaths in individuals less than 20 years old, respectively [4]. In general, survival from cancer has improved dramatically over the past forty years, presumably due to a combination of improved treatments and earlier detection. This is especially true in the paediatric population, among whom the overall long-term survival rate across all cancers has risen from under 50% to roughly 70% [4;8;9]. Even with brain cancer, the prognosis in children has improved, such that long-term survival is now achieved in more than half of patients [10]. This being said, the neoplasm originates in the brainstem in roughly 10-20% of children with CNS cancer [7;11-13], accounting for 150 to 300 new cases per year in the U.S. [11;13] and thus rendering it more common in children and adolescents than in adults [14;15]. nd in this subset of children, the prognosis generally is considered extremely bleak, akin to that of glioblastoma multiforme [1;4;14;16-19]. Despite a continued poor prognosis, much has changed over the past several decades like how paediatric brainstem tumours are diagnosed and classified, if and when surgery is considered, approaches to surgery, the use of adjunct therapies like radiation and chemotherapy, and the evolution of several new therapeutic options. Once lumped together as a single 2014 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License ( which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

3 32 Tumors of the Central Nervous System Primary and Secondary entity that was considered inoperable and, hence, completely incurable, recent developments in imaging and surgical techniques have led to the classification of several different types of brainstem tumour; and, for some of these, surgical resection is considered the treatment of choice [16;18;20]. This chapter reviews the historical progression of understanding of paediatric brainstem tumours, including evolving beliefs regarding their classification, diagnosis, management, prognosis and prognosticators, dating from the 1970s to current times. It then describes current diagnostic and management protocols for these tumours, ending with a glance forwards towards potentially promising treatments and technologies and how they might, hopefully within the near future, favourably alter outcomes in these patients, both in terms of their survival and quality of life. 2. Where we have been 2.1. Definitions, diagnosis and classification The brainstem has been defined as extending from the midbrain (tectal plate) to the medullary cervical junction [7]. rain stem tumours are, by definition, tumours that involve the brainstem. However, they include tumours not just in the brainstem per se, but also in the upper cervical spine [16]. In the paediatric population, posterior fossa tumours significantly outnumber those that are supratentorial [10;21;22], and brainstem tumours account for roughly 25% of all tumours found within the posterior fossa [16;23]. The vast majority of these are primary lesions, since just 3 to 5% of all brain metastases are found in the brainstem [24]. The most common paediatric posterior fossa tumours are cerebellar astrocytomas, medulloblastomas, ependymomas and brainstem gliomas [25]. Traditionally, the term brainstem glioma has been used to incorporate all brainstem tumours, largely because biopsies often were not performed and the majority of brainstem tumours in childhood are, in fact, of glial cell origin [26;27]. However, other histological forms of tumours do exist, though they comprise but a small percentage of brainstem tumours, and generally tend to be exophytic, growing either external to the brainstem or on its surface. Even among such exophytic tumours, gliomas form the clear majority. In their series of 75 paediatric patients with exophytic tumours seen between 1970 and 1990, Pierre-Kahn et al. [28] noted 69 glial tumours (92% of the total), of which 58 were astrocytomas and 11 oligodendrogliomas. The remaining six non-glial tumours were two ependymomas; two primitive neuroectodermal tumours (PNET); one ganglioglioma; and one of unknown histology [28]. In a much more recent review of non-glial brainstem tumours, other brainstem lesions noted to occur anecdotally included medulloblastomas invading the brainstem, cavernomas, lymphomas, haemangioblastomas, and other ganglionic and mixed tumors [29]. Nonetheless, many continue to use the term brainstem glioma generically, given that the vast majority of brainstem tumours are glial cell based. Up until the development of advanced imaging techniques like computed tomography (CT) and, to an even greater extent, magnetic resonance imaging (MRI) brainstem tumours

4 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 33 tended to be lumped together as a single clinical entity and considered uniformly inoperable [30]. Since they were presumed gliomas and surgery was deemed contraindicated, there was no call even to biopsy them, except in certain instances in which the diagnosis was in doubt. For example, in 1969, Matson wrote: brainstem gliomas must be classified as malignant tumors since their location in itself renders them inoperable [21]. nd as late as 1984, Tomita et al. [31] wrote: Since biopsy specimens often misrepresent the true pathology... surgery undertaken to obtain precise histological verification of brain stem gliomas is futile. Instead, these latter authors recommended computed tomography (CT) with high-resolution metrizamide CT cisternography to distinguish surgically resectable extra-axial tumors adjacent to the brain stem from non-resectable intrinsic brain stem gliomas. Such sentiments that tumours only were worth a biopsy if they were either completely exterior to brainstem tissue or, at worst, on the surface were echoed by others [32]. Consequently, early classification of brainstem tumours often subdivided lesions into those that were exclusively or primarily intraaxial, and those that were exclusively or primarily extra-axial. Meanwhile, brainstem tumours were collectively contrasted against those involving the midbrain or thalamus, both of which exhibited considerably superior prognoses [33]. This being said, there were early reports of certain patients with brainstem tumours who survived long-term. For example, in 1971, Lassiter et al. reported on 37 patients with presumed brainstem gliomas (22 of them children), among whom there were four children with large, surgically-drained neoplastic cysts who achieved long-term survival: two of these children died 7½ and 13 years later, and two still were alive 8½ and 9 years post diagnosis, one of them going on to graduate from college and the other with residual hemiparesis and mental retardation [34]. In 1975, Hara et al. reviewed the cases of 24 brainstem gliomas, and found that the median survival was 9 months, except for one patient who survived beyond 4½ years and another who lived for 14 years and 10 months before succumbing to the disease [35]. Similar isolated cases of long-term survival with brainstem gliomas, especially cystic lesions, were reported as far back as 1937 and 1940 through the mid nineteen seventies and early eighties [36-42]. nd, contrary to conventional wisdom, even then, some neurosurgeons were routinely performing biopsies and surgery on select patients with brainstem gliomas [27;40;41;43-46]. In the nineteen sixties and early nineteen seventies, angiography and pneumoencephalograms were largely used to identify brain and brainstem tumours [47-50]. These imaging modalities were replaced in the mid nineteen seventies as computed tomography (CT) emerged as the imaging tool of choice for the detection and diagnosis of both supra-and infra-tentorial lesions, including brainstem tumours [51-54]. y this point, some surgeons were starting to distinguish outcomes between different subsets of patient who were able to undergo either partial or radical resection of brainstem cancers [27;40;41;43-45]. The first to propose what he termed a staging system for brainstem gliomas was Epstein, in 1985, who first categorized tumours as intrinsic, exophytic and disseminated; and then subcategorized intrinsic (actual brainstem) tumours as either diffuse, focal or cervico-medullary [43]. In 1986 and again in 1988, Epstein et al. published series of paediatric patients undergoing surgery for brainstem cancer, again classifying patients as focal, diffuse or cervico-medullary; in the latter report, they added a

5 34 Tumors of the Central Nervous System Primary and Secondary category for cystic lesions [27;44]. What these authors noted was uniformly dismal outcomes in children with diffuse lesions, but often favourable outcomes in those patients with any one of the three other classifications. In the 1988 report, co-authored by Wisoff [27], among 66 children with intrinsic brainstem gliomas diagnosed between 1980 and 1986 who underwent radical surgical resection and either CT, MRI or both pre-operatively, 27 (41%) were found to have diffuse tumors, all of whom died within 12 to 18 months of surgery with malignant neoplasms that had not benefitted from surgery. However, five of nine children with cystic tumors, three of five with focal tumors, and twenty of twenty-four with cervicomedullary tumors were discovered to have a histopathologically low-grade lesion, and these 28 children all remained alive between one and six years post-operatively. The authors proposed criteria combining clinical and neuroradiological findings to predict which patients with brainstem tumors were likely to benefit from radical surgical intervention [27]. Since that time, several additional and very different classification systems have been proposed for brainstem tumours, including those by Stroink et al., arkovich et al., lbright, Fischbein et al., Choux et al., Fisher et al., Mehta et al. and, most recently, Ramos et al. [45;55-61]. (See Table 1) The system proposed by Stroink et al., published in 1987, was born out of a study of 16 children (9 girls, 7 boys; age range 1½-12 years) with dorsally exophytic transependymal benign brainstem gliomas diagnosed between 1962 and 1985 at Sick Kids Hospital in Toronto, Canada [45]. Of these, 13 were low-grade (grade I-II) astrocytomas, one a grade III astrocytoma, and two gangliogliomas. ll 16 patients underwent subtotal resection of their tumours, and seven had post-operative radiation therapy. One died 18 months after surgery, but the remaining 15 remained alive an average of eight years post-operatively (median=7 years; range=8 months to 23 years). ased upon these results, Stroink et al. proposed dorsal exophytic gliomas (Type I) as a distinct clinical entity, to differentiate them from hypo-dense, nonenhancing intrinsic tumours (Type IIa); hyper-dense, contrast-enhancing exophytic intrinsic tumours (Type IIb); focal, enhancing cystic intrinsic tumours (Type III); and focal iso-dense and contrast-enhancing intrinsic tumours (Type IV) [45]. Author(s) Year Published Imaging Classification System Epstein [43] 1985 CT, MRI Intrinsic - diffuse; focal, cervicomedullary Exophytic - anterolateral; posterolateral Disseminated - positive cytology; positive myelography Epstein et al. [44] 1986 CT, MRI Diffuse Focal Cervicomedullary Epstein et al. [27] 1988 CT, MRI Diffuse Focal Cervicomedullary Cystic

6 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 35 Author(s) Year Published Imaging Classification System Stroink et al. [45] 1987 CT with contrast Type I: dorsal exophytic glioma Type II intrinsic brainstem tumour * IIa - hypodense with no contrast enhancement IIb - hyperdense with contrast enhancement, exophytic Type III: focal cystic tumour with contrast enhancement Type IV: focal isodense tumour with contrast enhancement Barkovich et al. [55] 1990 MRI Tumours characterized by - (1) Location - midbrain, pons, medulla (2) Focality - focal or diffuse (3) Direction & extent of tumour growth (4) Degree of brainstem enlargement (5) Presence/absence of exophytic growth (6) Presence/absence of haemorrhage or necrosis (7) Evidence of hydrocephalus Albright [58] 1996 MRI Focal - midbrain; pons; medulla Diffuse Fischbein et al. [56] 1996 MRI Midbrain - diffuse; focal; tectal Pons - diffuse; focal Medulla - diffuse; focal; dorsal and exophytic Choux et al. [57] 2000 MRI Type I - diffuse Type II - intrinsic and focal Type III - intrinsic and exophytic Type IV - cervicomedullary Fisher et al. [59] 2000 MRI Pilocytic astrocytomas Fibrillary astrocytomas Other tumors Mehta et al. [60] 2009 MRI Intrinsic tumours - expanding; diffuse infiltrative; purely ventral Extrinsic tumours Ramos et al. [61] 2013 MRI Diffuse intrinsic/diffusely infiltrative Posterior exophytic/cervicomedullary gliomas; other focal tumours Dorsal exophytic tumours Cervicomedullary tumours Focal tectal tumours Table 1. Classification Systems for Brain Stem Tumours-1985 to Present

7 36 Tumors of the Central Nervous System Primary and Secondary It was in the mid to late nineteen-eighties that magnetic resonance imaging (MRI) started to replace CT as the imaging modality of choice for the diagnosis and classification of brainstem tumours [62]. s such, it was based upon MRI studies that arkovich et al. proposed their criteria for brainstem tumours in 1990 [55]. Their criteria were derived from the results of retrospectively-reviewed MRI studies for 87 paediatric patients with brainstem gliomas. T2- weighted images were deemed most appropriate for use, given that they were the most accurate at demonstrating the extent of tumour. In this much more elaborate classification scheme, tumours were characterized in terms of (1) their location of origin, into midbrain, pons and medulla; (2) their degree of focality (whether diffuse of focal); (3) the direction and extent of tumour growth; (4) the degree of brainstem enlargement; (5) the degree of exophytic growth; (6) the presence or absence of cysts, necrosis or hemorrhage; and (7) the presence or absence of hydrocephalus. This classification scheme was never scientifically validated to determine its ability to predict outcomes, however, and failed to achieve widespread acceptance. Likely reasons for this failure were how cumbersome the system was, and the lack of any firm guidelines as to its use. s such, subsequent classification schemes have been much simpler, starting with those proposed by lbright, in 1996, who returned to the very simple categorization of lesions as being either focal (further distinguished by location, into midbrain, pons or medulla) or diffuse [58]; and Fischbein et al., also in 1996, who again categorized lesions as focal or diffuse, but categorized both forms of lesion by location again into midbrain, pons or medulla [56]. Harkening back to the four-category system initially proposed by Epstein in 1985 [43], Freeman et al., in 1998, and Choux et al., in 2000, adopted the four relatively straight-forward categories largely in use today: diffuse intrinsic; focal intrinsic; focal exophytic, and cervicomedullary [18;57]. In 2009, Mehta et al. proposed a modified sub-categorization of intrinsic tumours, into expanding, diffuse infiltrative, and pure ventral varieties, achieving good surgical results and reasonable survival with the first of the three subtypes [60]. Then, most recently, Ramos et al. adopted yet another classification system that included diffuse intrinsic and diffusely infiltrative as a single category, followed by four additional categories of various focal lesions [61]. The authors conclusion was that brainstem tumors are a heterogeneous group of tumors. Clearly then, in terms of how brainstem tumours are now perceived, there has been a 180 degree reversal from early statements [21;30;31] about their homogeneity and the inappropriateness of biopsies for brainstem tumours. The questions remain, however: Who warrants such biopsies? nd how could biopsies and surgery influence outcomes? 2.2. Outcome predictors and indications for surgery Early universal pessimism regarding the fate of children with brainstem tumours has also clearly changed in recent years, with reports of better than 90% five-and ten-year survival rates in children with certain types of tumour [23;63]. The primary objective of all the classification schemes proposed to date has been to identify which patients warrant surgical treatment and how aggressive such treatment should be. Contrary to early years, when all brainstem tumours were considered inoperable [21;30;31], for the past thirty years neurosurgeons have been operating to either de-bulk or completely excise lesions when doing so was felt to be of clinical

8 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 37 benefit, potentially positively influencing quality of life and/or survival. s stated earlier, that a subset of patients existed for whom long-term survival was possible has been suspected for almost eighty years [36;37]. Once again, it was Epstein et al. who first identified diffuse lesions as being the most rapidly fatal, and focal and cystic lesions as being more amenable to treatment and long-term survival [27;43;44]. Since then, other prognostic factors have been identified that can influence the decision to operate (See Table 2). Factor Studies Year Published Imaging Comments Focal (vs. diffuse intrinsic) tumour Sandri et al. [77] 2006 MRI 87.4% 4-year vs. 12.3% 2-year survival Mauffrey [20] 2006 MRI 90% vs. 22% 2-year survival Fried et al MRI 89% vs. 3% 5-year survival Cystic appearance Lassiter et al. [34] 1971 not identified All 4 children with cystic lesions Prolonged pre-diagnosis symptoms Fisher et al. [59] 2000 MRI Cox regression survival w/ symptoms > 6 mo (p =. 004) Shuper et al. [17] 1998 MRI Mean survival 19.5 vs mo with >1 mo symptoms Other than pons location Fisher et al. [59] 2000 MRI Cox regression survival w/ pons involvement (p=. 0002) No presenting eye symptoms/eye palsies Fisher et al. [59] 2000 MRI Cox regression survival w/ abducens palsy (p<.0001) Shuper et al. [17] 1998 MRI Mean survival 17.5 vs mo with no eye symptoms Basilar artery engulfment Fisher et al. [59] 2000 MRI Cox regression survival w/ basilar artery engulfed (p=.006) Tectal location Poussaint et al. [65] 1998 MRI All 32 children with tectal tumours alive after a mean 5 years follow-up Non-enhancing on gadolinium MRI Poussaint et al. [65] 1998 MRI Gadolinium enhancement of tectal lesions increased odds of disease progression x15 Contrast enhancement on MRI Dellaretti et al. [64] 2012 MRI Mean survival with contrast enhancing lesions = 21.7 mo.; with non-enhancing lesions 54.2 months (p <.001) Maximum diameter < 0.5 cm Poussaint et al. [65] 1998 MRI Each 1 cm increase in maximum diameter increased odds of disease progression x5 Neurofibromatosis type 1 Fried et al. [72] 2012 MRI All 7 children with NF-1 had non-focal lesions & long-term survival Other papers Numerous isolated anecdotal reports in other papers Table 2. Factors Predicting Favourable Outcomes in Paediatric Brainstem Tumour Patients

9 38 Tumors of the Central Nervous System Primary and Secondary One of the earliest attempts to empirically look at prognostic factors beyond imaging results was published by Fisher et al. in 2000 [59]. They reported on the results of a study of 77 patients, 21-years old or younger, seen between 1980 and In this study, they sought to identify characteristics statistically associated with poor survival. The factors that they identified were (a) symptom duration less than six months before diagnosis (p=0.004); (b) abducens palsy at presentation (p < ); (c) a pontine location (p=0.0002); and (d) engulfment of the basilar artery (p=0.006). Twenty of their patients were found to have pilocytic astrocytomas, which were associated with a very favorable 5-year overall survival rate of 95%, as well as with location outside the ventral pons (p=0.001) and dorsal exophytic growth (p=0.013). nother histological type they could predict based upon clinic-radiographic findings was fibrillary astrocytoma, of which they had 14 cases. These lesions were associated with symptoms < 6 months (p=0. 006), abducens palsy (p < 0.001), engulfment of the basilar artery (p=0.002), and a one-year survival rate of just 23% (p < ). Shuper et al. also found, in their analysis of 24 children operated upon between 1981 and 1997, that a shorter duration of symptoms (< 4 weeks) and visual symptoms at presentation were associated with shorter survival and lower survival rates [17]. Meanwhile, Dellaretti et al. found that contrast-enhancing lesions on MRI were associated with a significantly shorter mean duration of survival versus non-enhancing lesions (21.7 vs months, p < 0.001); however, on Cox proportional hazards regression analysis, tumour grade was the only significant predictor of survival, suggesting that contrast enhancement in that sample was an indicator of higher tumour grade [64]. Poussaint et al. specifically sought to determine which clinical and imaging findings best correlated with outcomes in children with tumours involving the midbrain tectum, via a retrospective review of the medical records and imaging studies of 32 children (16 boys and 16 girls; mean age, 8 years) with tectal tumours [65]. Of this number, eight children had undergone CT, 11 MRI, and 13 both CT and MRI studies. Over a mean follow-up period of five years (range, 3.6 months to 17 years), all patients experienced hydrocephalus, for which all but one required cerebrospinal fluid (CSF) diversion. The tectum was the centre of the tumour in all cases; and the majority of the tumours appeared iso-dense on CT scans, isointense on T1-weighted MR images, and hyper-intense on T2-weighted images. Twenty patients required no further treatment. In this group, the mean maximum tumour diameter was 1.8 cm, and enhancement occurred in only two cases (10%). t follow-up, 18 tumours were the same size as at baseline, one was larger due to cyst formation in the setting of stable symptoms, and one was smaller. The remaining 12 patients required further treatment (excision and/or radiotherapy) because of disease progression, indicated either by increased tumour size or by worsening symptoms. In this group, the mean maximum tumour diameter was 2.5 cm and contrast enhancement occurred in nine (75%). Further follow-up in this group showed decreased tumour size in eight and stable residual tumor in three. The authors then used regression analysis to calculate and compare the likelihood of a patient requiring further treatment with various-size enhancing versus non-enhancing tumours and identified two trends: larger tumours were more likely to require further treatment; and the same was true for enhancing lesions. Combining these two factors was especially predictive. For example, whereas 31% of lesions with a maximum diameter of 0.5 cm required further treatment, samesized non-enhancing lesions warranted such treatment less than one percent of the time.

10 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 39 Corresponding percentages for lesions of maximum diameter 1.0, 1.5, 2.0, 2.5, 3.0, 3.5 and 4.0 cm were 48 and 1%, 67 and 3%, 80 and 6%, 90 and 11%, 95 and 21%, 97 and 36%, and 98 and 54%. y the time a lesion reached 4.5 cm in maximum diameter, virtually all (99 and 70%) required further treatment, irrespective of whether they did or did not enhance. Overall, the odds of surgical or radiation treatment were almost five times greater for each 1 cm increase in maximum tumor diameter (odds ratio, 4.9; 95% confidence interval, ; p=0.015); and 15 times greater when the tumor enhanced versus when it did not (15.0; ; p < 0.003). The investigators thereby concluded that, though paediatric tectal tumors exhibit somewhat variable behavior, patients generally do well, and that larger tumours, and especially those that that enhance with contrast on MRI, are highly likely to be more aggressive [65]. Whether size and contrast-based enhancement on MRI or CT predict the need for treatment or ultimate outcomes with other non-diffuse intrinsic or exophytic brainstem tumours has not yet been demonstrated. nother factor long believed to place individuals at increased risk of brainstem and other CNS malignancies, but also to confer a relatively favorable prognosis, is the presence of type 1 neurofibromatosis (NF-1), though this belief is largely based upon anecdotal reports [66-72]. In the large study of 223 children with brainstem tumours reported by Fried et al., however, there were seven children with concomitant NF-1; and all had low-grade brainstem tumours [72]; the statistical likelihood that this occurred merely by chance, assuming a fifty-fifty split of low-to high-grade lesions (which is roughly what they observed across the sample), is less than one percent. Overall in NF-1 patients, brainstem gliomas comprise a heterogeneous group of lesions, consisting of three main subtypes: (1) diffuse brainstem enlargement; (2) focal enhancing nodules with or without cystic areas; and (3) peri-aqueductal gliomas. ll of these subtypes, including diffuse brainstem enlargement, generally exhibit a very indolent course and do not require treatment, though MRI monitoring is indicated until their indolent course is confirmed. Some lesions even regress on their own [73]. The diffuse brainstem enlargement is somewhat similar in appearance to the unidentified bright objects (U Os) that are the most commonly observed CNS lesions on MRI in patients with NF-1 [74]. Like U Os, they exhibit abnormal signals on T1-weighted images. The major differences are that, as opposed to U Os, there usually is a mass effect, and these lesions also tend to be considerably larger than most U Os. What the diffuse enlargement represents remains controversial. Many presume them to be gliomas, but they exhibit a much more indolent course than brainstem gliomas seen outside of NF-1, such that adjuvant treatment is only required in the minority of patients whose lesions progress. However, ongoing monitoring is required to detect the few who do progress, before neurological deficits ensue, which often are irreversible. Rarely, these gliomas progress to more malignant forms of astrocytoma, including glioblastoma [73;75]. The focal enhancing nodules seen in the brainstem of NF-1 patients, which may occur with or without cystic areas, generally are thought to represent pilocytic astrocytomas, given their imaging characteristics. Like pilocytic astrocytomas elsewhere, they generally are indolent; but their course is unpredictable and the brainstem so susceptible to major deficits, relative to the cerebral hemispheres, that ongoing monitoring is required. Small, focal intrinsic lesions

11 40 Tumors of the Central Nervous System Primary and Secondary may enlarge and then regress spontaneously. Exophytic tumours often are more aggressive and require treatment. Periaqueductal gliomas occur adjacent to the aqueduct of Sylvius between the 3 rd and 4 th ventricles in the midbrain. They typically manifest with late-onset aqueductal stenosis, leading to hydrocephalus. Presumably, they represent low-grade gliomas or glial hamartomas, and typically are indolent. However, because of their location, CSF shunting often is necessary. ut, as with the other two forms of brainstem tumour seen in NF-1 patients, resection usually is unnecessary [73;75]. Table 3 summarizes currently published research on tumour characteristics that predict a favourable outcome. What is most evident is the dearth of confirmatory studies. These findings aside, the brainstem tumour characteristic that is unquestionably the one most likely to dissuade surgeons from operating is evidence that the tumour is intrinsic and diffuse, due both to the known high risks of surgery and to the lack of evidence suggesting any benefit, in terms of either quality of life or survival time [16-18;20]. The corollary to this is that, contrary to thirty years ago when most surgeons preferred to leave all brainstem tumours alone, focal, exophytic and cystic tumours are increasingly being accessed and resected, often with good results [16-18;20;23;60]. Year Study Study subjects 5-year survival Author(s) Published Type N = N = % = Lesniak et al. [76] 2003 Retrospective % Sandri et al. [77] 2006 Retrospective % Mauffrey [20] 2006 Retrospective % Teo et al. [63] 208 Retrospective % Fried et al. [72] 2012 Retrospective % Klimo et al. [23] 2013 Retrospective % OVERALL % Table 3. Survival in Paediatric Patients with Non-Diffuse Intrinsic Brainstem Tumours 2.3. Treatment and outcomes In 1984, Tomita [31] wrote that radiation therapy is the choice of treatment, should CT indicate clear evidence of intrinsic brain stem tumor... posterior fossa craniotomy should be undertaken only for aspiration of cystic intrinsic stem tumors, resection of extra-axial juxtastem tumors and, although rare, in instances when CT is unable to definitively distinguish extra-axial from intra-axial mass for verification of lesion location. Clearly, sentiments have changed; but has the increase in surgical interventions altered outcomes? In 1998, Shuper et al. published the results of their study of 24 children with brainstem tumours operated upon between 1981 and 1997 [17]. The main question they asked was: are we

12 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 41 improving outcomes? tissue diagnosis was achieved in only six of the children. lthough the investigators did not perform inferential statistical analysis, average survival in five patients seen before 1990 was 8.6 months, versus 19.4 months in the five patients seen in 1990 through 1994, and 20.0 months in those seen in 1995 and Moreover, raw data were provided, allowing the current author to perform a statistical comparison of months of survival pre (8.6) versus post (19.7) January 1 st This difference, despite the low numbers, is significant (t=2.379, df 13, p=0.03). However, these patients generally were offered radiation +/-chemotherapy, and not surgery. In addition, other biases might have erroneously generated these results, like alterations in referral patterns or the earlier recognition of tumours as a result of major technological advances in imaging. More recently, several authors have reported survival rates in paediatric patients treated surgically for non-diffuse tumours (Table 3). In 2003, Lesniak et al. [76] retrospectively reviewed the charts of all pediatric patients admitted to Johns Hopkins University Hospital with a diagnosis of a brainstem tumor between January 1985 and December 2000: 89 patients met the inclusion criteria, among whom 57 (64.0%) underwent surgical resection, while 32 (36%) were treated with radiation and/or chemotherapy. Of the surgical candidates, 57 (100%) had an accompanying MRI scan significant for an enhancing lesion in the midbrain, pons or the medulla. The pathology was consistent with juvenile pilocytic astrocytoma in 30 patients (52.6%) and glioblastoma multiforme in 12 patients (21.1%). The remaining cases consisted of ten patients (17.5%) with fibrillary astrocytomas, three (5.3%) with gangliogliomas, one (1.8%) with an oligodendroglioma and one (1.8%) with a primitive neuroectodermal tumor. Total surgical resection was attained in 29 patients, near total resection (>90%) in eight, subtotal resection (50-90%) in 15, and partial resection (<50%) in five. The progression-free survival of all patients, which included the twelve with glioblastomas, was 71.9% at 3 years and 45.6% at 5 years. Excluding the 12 patients with known glioblastomas, all of whom died prior to three years, these survival rates rise to 83 and 53%, comparable to rates observed in a subsequent study by Sandri et al. [77]. In 2006, Sandri et al. [77] published their series of 31 children admitted to their institution from 1995 to 2003, 14 of whom were classified on MRI as having diffuse and 17 as having focal brainstem tumours. Patients with diffuse lesions were treated with locoregional radiotherapy (1.8 Gy/day for 54 Gy) and weekly vincristine for radiosensitization (1.5 mg/sm for six total doses). Meanwhile, patients with focal tumours underwent surgical resection, with adjunct chemotherapy and/or radiotherapy considered on a case-by-case basis. mong the 14 with diffuse tumours, ten experienced a partial response, three exhibited stabilization of their disease, and one progressed. General and/or neurological symptoms improved in more than 80% of these patients. However, the median time from diagnosis to progression and from diagnosis to death were just 8 (range of 3-13) and 13 (range of 4-25) months, respectively, with a 2-year overall survival rate of just 12.3%. Conversely, among the 17 children with focal lesions, gross total removal was achieved in 4/17 cases, subtotal removal in 7/17, and partial removal in 6/17. There was one surgery-related death. Eight out of 17 patients had adjuvant chemo-and/or radiotherapy after progression, among whom six remained free of neurological symptoms and two died secondary to tumor progression. The 4-year overall and disease-free

13 42 Tumors of the Central Nervous System Primary and Secondary survival rates were 87.4 (SE 8.4) and 58.8% (SE 11.9), respectively, with the extent of resection identified as the best predictor of survival (p=0.012) [77]. lso in 2006, Mauffrey reported on his retrospectively reviewed series of 27 paediatric patients admitted to hospital in Turin, Italy with a diagnosis of brainstem glioma [20]. Thirteen patients had a diffuse pontine tumour on MRI scan, while fourteen had other brainstem gliomas. Those in the first group had a shorter mean duration of symptoms prior to diagnosis (2.6 vs months), never demonstrated gadolinium enhancement of their tumour on MRI (vs. 78.6% in the other group), and were much more likely to have symptoms or findings indicating cranial nerve involvement (77.0 vs. 28.5%). None of the 13 with diffuse gliomas underwent radical surgery, whereas it was the treatment of choice in the remaining 14. Two-year survival rates were 25% and 90%, respectively, and 60% of the latter remained alive at five years [20]. In 2008, Teo and Siu [63] reported on their results with 34 consecutive patients between 3 and 16 years of age who underwent endoscope-assisted microsurgery for focal brainstem gliomas with the intent of radial resection between 1999 and More than 90% tumour resection was achieved in 31patients, while >50% was attained in the remainder. There was no perioperative mortality and the average follow-up was 46 months. Twenty-three patients (74%) harboured low-grade and 11 (26%) high-grade gliomas. Kaplan-Meier survival analysis revealed marked differences in the 5-year survival rates between the two groups (100% vs. 33%). Multivariate analysis demonstrated that the degree of tumour resection was not associated with poor outcome at 6 months. Two papers published over the past couple of years are those by Fried et al. [72], based at The Hospital for Sick Children in Toronto Canada, and Klimo et al., based at George Washington University in Washington, DC [23]. The latter study, like all those described previously, was relatively small, with just 52 patients (32 boys), all with radiographically-confirmed, low-grade focal brainstem gliomas seen from 1986 to The median duration of follow-up was 10.0 years, and the median age at diagnosis 6.5 years (range 1-17 years). Tumors were located in the midbrain (n=22, 42%), pons (n=15, 29%), and medulla (n=15, 29%). Surgical extirpation was the primary treatment in 25 patients (48%). Five-and 10-year event-free survival and overall survival rates were 59% and 98%, and 52% and 90%, respectively. Surprisingly, children with intrinsic tumors trended towards slightly higher event-free survival at 5 years than those with exophytic tumors (p=0.054), but not at 10 years (p=0.147). No other variables were predictive of event-free survival [23]. The retrospective study by Fried et al. is by far the largest to date, assessing a total of 223 children with brainstem tumours (12% of all CNS tumours seen) followed at The Hospital for Sick Children in Toronto over the preceding 25 years [72]. Ninety-five of these tumours were diffuse and an additional 17 were high-grade astrocytomas (grade III or IV). The investigators made several novel observations. First, whereas 75% of tumors involving the pons were highgrade, 98% of tumours lacking pontine involvement were low-grade (p = ). Second, residual tumour after surgery, even when visualized, did not adversely alter either progression-free survival or overall survival. nd, among those requiring further treatment, 5-year progression-free and overall survival were comparable between those receiving chemotherapy (53 and 93%) and those administered radiotherapy (66 and 83%) (p = 0.26 and 0.30, respective

14 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 43 ly). mong those with focal lesions, five-year progression-free survival (PFS) and overall survival (OS) were 57% and 89%. Combining the results of these last seven studies (Table 3), in which children with non-focal brainstem tumours were identified and treated, either surgically or non-surgically, there were 282 such children, of whom 223 (79.1%) survived long-term (beyond 4 years and most beyond 5 years). In the largest study, which had 108 such children [72], 90% remained alive 10 years post diagnosis. Several comments are warranted. First, every one of the above-mentioned studies had significant methodological issues, not the least being that they all were retrospective. Moreover, because patients were selected for surgery, it is not possible to rule out the possibility that the characteristics that made them surgical candidates in the first place (e.g., a focal vs. diffuse tumour) were responsible for their survival, rather than the surgery itself. Nonetheless, what is clear is that long-term survival among children with non-diffuse tumours clearly is not at all uncommon. In fact, it seems likely for four out of five such children, making it the rule and not the exception, which is radically distant from both the beliefs and outcomes of thirty and forty years ago, and considerably improved from survival rates in at least one older study in which those with non-diffuse tumours were identified but not offered surgery [78]. Fried et al. concluded that children with non-diffuse brainstem tumours do well even with conservative management; but their conservative management appears to have consisted of surgical resection in most cases, with adjuvant chemotherapy or radiation considered more aggressive [72]. Ideally, from a purely methodological standpoint, future studies would randomize surgery-eligible patients to surgery versus non-surgery groups. Practically, however, this author considers it unlikely that most parents would consent to allow their children to be randomly assigned to one group or the other when surgery offers at least the potential of cure. The good news, then, irrespective of methodological questions regarding the true effectiveness of surgery, is that children with several types of brainstem tumour appear to do well. The bad news is that the majority of brainstem tumours in children continue to be either diffuse intrinsic lesions or high-grade astrocytomas [72], and the prognosis for either remains dismal, with only about 10% of children alive two years after diagnosis [79;80]. One major challenge currently facing researchers and clinicians is to identify non-surgical ways to better target these otherwise poorly responsive tumours. 3. Where we are and where we are going Clearly, the past forty years have brought about substantial changes in the way brainstem tumours are both perceived and managed. No longer is the term brainstem glioma used to lump all brainstem lesions together as homogenous, untreatable and ultimately fatal. Now, either three or four heterogeneous classes of tumour are described, and all but one is considered to be an indication for aggressive, often surgical treatment. Some investigators sub-classify lesions further, related to their anatomical location (e.g., midbrain, pons, medulla). In essence, though, irrespective of the number of tumour types proposed, all categorization systems ultimately separate diffuse intrinsic lesions from all others.

15 44 Tumors of the Central Nervous System Primary and Secondary Current diagnosis largely depends on the use of patient histories and physical findings, especially paying attention to the duration of symptoms prior to diagnosis and the type(s) of neurological deficits identified (e.g., ocular palsies), followed by magnetic resonance imaging, both with and without contrast. Lesions that are intrinsic and diffuse, and lesions that involve the pons are generally considered to be a contraindication to surgery, though many clinicians are now biopsying them to determine if they are high-or low-grade to estimate course and likely survival time [81]. Conversely, lesions that are focal, cystic, and extraphytic are usually considered treatable, with surgery to remove as much of the tumour as possible often considered the treatment of choice. Table 4, utilizing the most recently proposed classification system of Ramos et al. [61], summarizes current general practices. Following the 2013 publication of guidelines promoting the biopsy of diffuse intrinsic brainstem lesions to aid in the development of targeted therapies [82], stereotactic biopsies must now be at least considered standard of care for these patients. Radiation therapy, sometimes combined with chemotherapy, appears to induce brief clinical remissions in patients with diffuse tumours for whom surgery is not indicated, and may prolong survival slightly [83]. With other tumours, surgery often plays the major role; though some patients merely require monitoring and others respond well to radiation or conservative measures like CSF shunting alone. Tumour type Typical location Usual histology Biopsy Surgery Radiatio Chemothera n py 2-year survival Diffuse Intrinsic or Pons grade III-IV gliomas Yes No Yes +/- 10% Diffusely fibrillary Infiltrative astrocytomas Focal Medulla pilocystic astrocytomas Yes Complete resection +/- +/- 90% Brainstem gangliocytomas often possible Dorsal Exophytic Mostly in 4th pilocystic astrocytomas Yes Usually only partial +/- +/- 90% ventricle other low-grade glioma resection is possible Cervicomedullary Epicentre either in medulla or cervical spine pilocystic astrocytomas gangliocytomas Yes Complete resection +/- +/- ~ 100% possible in ~75% of pts Focal Tectal Tectum low-grade gliomas Yes Shunt placement Yes Usually not ~ 100% Resection often unnecessary Table 4. Current Management of Paediatric Brainstem Tumours

16 Paediatric Brainstem Cancers Where We Have Been; Where We Are; Where We Are Going 45 There remains no consensus, however, as to how to best categorize lesions; and the lion s share of brainstem tumours continue to be considered untreatable and to have a dismal prognosis. Consequently, three major challenges that remain are (1) coming to some consensus as to classifying lesions, so as to best predict their course, likely response to treatment and, hence, when and how best to treat them; (2) developing more effective non-surgical treatments to treat the intrinsic diffuse tumours for which all current treatments, including surgery, have been ineffective and the prognosis remains abysmal; and (3) optimizing quality of life in these children and their families, irrespective of long-term prognosis. With a view to these three main objectives, we now briefly explore current and future directions in the detection, diagnosis and classification of brainstem tumours, and in their non-operative and operative management Enhanced diagnostics and classification In few fields have there been as many and as dramatic advances as in the field of diagnostic imaging, and this appears to be having a significant impact upon how brain and brainstem tumours are now detected and diagnosed. Increasing recognition that not all brainstem tumours are a diverse collection of pathological entities with distinct courses and responses to treatment, has led to attempts to distinguish between them using advanced imaging. Typically, for example, diffuse intrinsic fibrillary astrocytomas appear hypo-intense on T1-weighted images, while heterogeneously hyper-intense on T2-weighted images; they also exhibit indistinct margins that reflect the tumour s highly infiltrative nature [61]. eyond distinctions made using different MR sequences, various contrasts are now being used to try to detect lesions that otherwise might be missed [84] and to delineate low-from high-grade lesions [19; 45;56;64;76]. For example, dorsal exophytic tumours often are low-grade pilocytic astrocytomas that, like high-grade fibrillary astrocytomas, may be hypo-intense on T1-weighted images and hyper-intense on T2-weighted images; however, they classically appear well-demarcated. Moreover, after gadolinium infusion, a cystic component often is identified, as only the solid portion of the tumour is enhanced, revealing a hypo-intense centre [61;85;86]. More recent advances in MR scanning like MR spectroscopy, MR perfusion, and diffusion tensor imaging (DTI) are also being utilized to further establish the histopathologic diagnosis of brainstem lesions [18;61]. n additional advantage of these newer MR technologies is that they are better at monitoring for disease recurrence or progression after treatment, since radiationinduced necrosis may be mistaken for tumour re-growth with traditional MRI [87-90]. In addition, functional scans like functional MRI (fmri), positron emission tomography (PET), and single-photon emission computed tomography (SPECT) are emerging as additional imaging tools to identify and characterize lesions in the brain, brainstem and spinal cord [19;91-95]. Further development of these tools may aid in further delineating the various brainstem tumours, even obviating the need for tissue in those patients for whom biopsies pose undue risk. This being said, among the most major imaging advances related to brainstem tumours relates to stereotactic biopsies. stereotactic biopsy utilizes a computer and images performed in at least two planes, first to localize a target lesion, like a tumour, in three-dimensional space; then

17 46 Tumors of the Central Nervous System Primary and Secondary to determine its depth; and finally to guide the removal of tissue for pathological examination. Stereotactic biopsies rely on the underlying principle of parallax, a process that initially was used in astrology to estimate distances between stars [96]. Parallax is the concept of using multiple site lines to visualize the same object relative to objects of known position in front of and behind it. Since objects closer to the observer tend to move more than objects that are more distant when the point of observation changes, measuring the degree of movement of the target lesion relative to reference points, combined with recording the change in viewing angle, permits one to estimate the depth or "Z-dimension" of the target lesion [97]. Long used to aid in the biopsy of breast lesions [98], CT-guided stereotactic biopsies have been used for the diagnosis of brain lesions since the nineteen seventies [99;100], but not with brainstem lesions until more recently. Just this year, Cage et al. reported on their results specifically biopsying diffuse brainstem tumours; they also extensively reviewed the literature [81]. In their own series, nine children with pontine lesions were biopsied, with successful tissue collection achieved in all cases. mong these lesions, four were found to be low-grade (grade I or II) astrocytomas, and five high-grade (grade III or IV) gliomas, demonstrating heterogeneity even among diffuse pontine lesions. Moreover, only one patient experienced any post-operative complication transient seizures and hydrocephalus. In their review of the literature, the same authors identified twenty case series besides their own, ranging in size from a single paediatric patient [101] to 52 children [102]. In this latter series, among 52 paediatric brainstem biopsies, only five patients experienced any post-operative morbidity, in four instances transient; and there were no deaths. iopsy was felt to alter management in 18% of the cases. In another one of the larger series, Kondziolka and Lunsford reported on their use of CT-guided stereotactic biopsies in 40 consecutive patients seen over a 13-year interval [103]. Of this number, 20 patients had midbrain lesions (n=20), 18 pontine lesions, and two medullary lesions. Midline lesions were approached via a coronal, trans-thalamic trajectory; lateral brain stem lesions usually were approached via a trans-cerebellar route. histologic diagnosis was achieved in 38 patients (95%), and no post-biopsy haemorrhages were noted on CT performed immediately after the procedure. Only one patient (2.5%) experienced a complication, which was transient diplopia. ltogether across the twenty-one papers Cage et al. reviewed, there were 294 documented brainstem biopsies in paediatric patients, among whom there were 16 cases of an intra-operative complication (5.4%); 42 cases of increased post-operative morbidity (14.3%) that was transient in the vast majority of cases; 29 inconclusive biopsies (9.9%); and two procedure-related deaths (0.7%) [102]. The authors concluded that brainstem biopsies are safer than widely perceived and that, when used judiciously, might be both safe and of advantage in terms of determining treatment. Stereotactic biopsies have been demonstrated to be superior to MRI alone in accurately diagnosing brainstem lesions [104] Enhanced non-operative management Many have argued against biopsying diffuse intrinsic brainstem lesions, since they cannot be surgically removed or even de-bulked without causing unacceptable neurological deficits because of their location and infiltrating nature, and given the lack of enduring response that the vast majority of patients demonstrate to traditional chemotherapy and radiation. ut this pessimistic outlook may be changing, given emerging knowledge about potential therapeutic